Information processing device, information processing method, and program

Abstract

To implement a highly accurate business support technique.SOLUTION: An information processing device includes: first acquisition means that obtains both an instruction content and reference information which is information to be referred to in processing indicated by the instruction content; generation means that generates a prompt by referring to the instruction content and the reference information; and second acquisition means that obtains a generated sentence which is generated based on the prompt.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] Technologies are being developed to support various tasks such as document creation and confirmation. For example, Patent Literature 1 discloses a technology for checking contracts using a neural network that has been trained in advance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-166864 A Summary of the Invention [Problem to be solved by the invention]

[0004] Generally, accuracy is required in tasks such as creating and checking documents. However, the technology described in Patent Document 1 has a problem in terms of accuracy.

[0005] The present disclosure has been made in consideration of the above problems, and has an object to provide a business support technology with high accuracy. [Means for solving the problem]

[0006] An information processing device according to one aspect of the present disclosure includes a first acquisition means for acquiring instruction content and reference information which is information to be referred to in the processing indicated by the instruction content, a generation means for generating a prompt by referring to the instruction content and the reference information, and a second acquisition means for acquiring a generated sentence generated based on the prompt.

[0007] An information processing method according to one aspect of the present disclosure includes at least one processor acquiring instruction content and reference information that is information to be referred to in a process indicated by the instruction content, generating a prompt by referring to the instruction content and the reference information, and acquiring a generated sentence generated based on the prompt.

[0008] A program relating to one aspect of the present disclosure causes a computer to execute a first acquisition process to acquire instruction content and reference information that is information to be referred to in the processing indicated by the instruction content, a generation process to generate a prompt by referring to the instruction content and the reference information, and a second acquisition process to acquire a generated sentence generated based on the prompt. Effect of the Invention

[0009] According to the present disclosure, a highly accurate business support technology can be provided. [Brief description of the drawings]

[0010] [Figure 1] 1 is a block diagram showing a configuration of an information processing device according to the present disclosure. [Diagram 2] 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Diagram 3] 1 is a block diagram showing a configuration of an information processing system according to the present disclosure. [Figure 4] 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Diagram 5] 11 is a diagram for explaining a process for generating a prompt and obtaining a generated sentence according to the present disclosure. FIG. [Figure 6] 11 is a diagram for explaining a process for generating a prompt and obtaining a generated sentence according to the present disclosure. FIG. [Figure 7] 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Figure 8] 11 is a diagram for explaining a process for generating a prompt and obtaining a generated sentence according to the present disclosure. FIG. [Figure 9]11 is a diagram for explaining a process for generating a prompt and obtaining a generated sentence according to the present disclosure. FIG. [Figure 10] 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Figure 11] 11 is a diagram for explaining a process for generating a prompt and obtaining a generated sentence according to the present disclosure. FIG. [Figure 12] 11 is a diagram for explaining a process for generating a prompt and obtaining a generated sentence according to the present disclosure. FIG. [Figure 13] 1 is a block diagram showing a configuration of an information processing system according to the present disclosure. [Figure 14] 1 is a flow diagram showing the flow of an information processing method according to the present disclosure. [Figure 15] 1 is a block diagram showing a hardware configuration of an information processing device according to the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] The following are examples of embodiments of the present invention. However, the present invention is not limited to the embodiments exemplified below, and various modifications are possible within the scope of the claims. For example, an embodiment obtained by appropriately combining the technical means employed in each of the embodiments exemplified below may also be included in the scope of the present invention. Alternatively, an embodiment obtained by appropriately omitting a part of the technical means employed in the embodiments exemplified below may also be included in the scope of the present invention. In addition, the effects mentioned in each of the exemplary embodiments shown below are examples of effects expected in the exemplary embodiments, and do not define the scope of the present invention. In other words, an embodiment that does not exhibit the effects mentioned in each of the exemplary embodiments shown below may also be included in the scope of the present invention.

[0012] [First embodiment] A first exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. This exemplary embodiment is the basic form of each exemplary embodiment described later. The scope of application of each technical means adopted in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technical means adopted in this exemplary embodiment can be adopted in other exemplary embodiments included in this disclosure to the extent that no particular technical obstacle occurs. In addition, each technical means shown in the drawings referred to for explaining this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure to the extent that no particular technical obstacle occurs.

[0013] (Configuration of information processing device 1) The configuration of an information processing device 1 according to this exemplary embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the information processing device 1. As shown in Fig. 1, the information processing device 1 includes a first acquisition unit 11, a generation unit 12, and a second acquisition unit 13.

[0014] (First Acquisition Unit 11) The first acquisition unit 11 acquires the instruction content and reference information to be referred to in the process indicated by the instruction content, and provides the acquired instruction content and reference information to the generation unit 12 described later. As an example, the first acquisition unit 11 acquires an instruction from a user, and identifies the instruction content by referring to a wording included in the instruction. Alternatively, the first acquisition unit 11 may be configured to acquire information indicating a user's selection regarding the instruction, and identify the instruction content by referring to the information.

[0015] The reference information acquired by the first acquisition unit 11 may be, for example, information selected so as to improve the accuracy of the sentence generated by the language model that references the prompt generated by the generation unit 12 described later. However, this does not limit the present exemplary embodiment. Furthermore, the expression "information to be referenced" in the above description does not limit the present exemplary embodiment, and may simply be expressed as "information to be referenced." Furthermore, the specific contents of the instruction and the reference information also do not limit the present exemplary embodiment, but the following is an example.

[0016] (Example 1) The instructions include an instruction to evaluate the evaluation target, The reference information includes past evaluation information that is information indicating past evaluations of the evaluation target.

[0017] (Example 2) The instruction content includes an instruction to analyze the analysis target, The reference information includes first type viewpoint information indicating one or more analysis viewpoints in the analysis process.

[0018] (Example 3) The instruction content includes an instruction to check a check target, The reference information includes comparison information that is information to be compared with the check target in the check process. (Generation unit 12) The generation unit 12 generates a prompt by referring to the instruction content and the reference information acquired by the first acquisition unit 11. Here, the prompt generated by the generation unit 12 is a prompt to be input to a language model described later, and the language model generates a generated sentence described later by referring to the content of the prompt. Note that the specific configuration of the prompt generated by the generation unit 12 does not limit this exemplary embodiment, but as an example, the generation unit 12 may be configured to generate the prompt including the instruction content and the reference information. Also, the prompt generated by the generation unit 12 is a prompt including natural sentences, but this does not limit this exemplary embodiment.

[0019] (Second Acquisition Unit 13) The second acquisition unit 13 acquires the generated sentence generated based on the prompt generated by the generation unit 12. More specifically, the second acquisition unit 13 inputs the prompt generated by the generation unit 12 into a language model, and acquires the sentence generated by the language model (also simply referred to as a "generated sentence"). Here, the language model may be a language model executed inside the information processing device 1, or a language model executed by a generation device communicably connected to the information processing device 1. Note that the language model is, as an example, a language model that refers to a prompt including a natural sentence and generates a generated sentence as a natural sentence, and is a language model that is machine-learned in advance, but this does not limit the present exemplary embodiment.

[0020] (Effects of information processing device 1) As described above, the information processing device 1 is configured to obtain an instruction and reference information that is information to be referred to in the process indicated by the instruction, generate a prompt by referring to the instruction and the reference information, and obtain a sentence generated based on the prompt. Therefore, the information processing device 1 can provide a highly accurate task support technology.

[0021] (Flow of information processing method S1) Next, the flow of the information processing method S1 according to this exemplary embodiment will be described with reference to Fig. 2. Fig. 2 is a flow diagram showing the flow of the information processing method S1. As shown in Fig. 2, the information processing method S1 includes a step (process) S11 of acquiring instruction content and reference information, a step (process) S12 of generating a prompt, and a step (process) S13 of acquiring a generated sentence.

[0022] (Step S11) In step S11, the first acquisition unit 11 acquires the instruction content and reference information that is information to be referred to in the process indicated by the instruction content, and provides the acquired instruction content and reference information to the generation unit 12. A more specific description of the first acquisition unit 11 has been given above, and therefore will not be repeated here.

[0023] (Step S12) In step S12, generation unit 12 generates a prompt by referring to the instruction content and the reference information acquired by first acquisition unit 11. Since the specific process related to generation unit 12 has been described above, a description thereof will be omitted here.

[0024] (Step S13) In step S13, the second acquisition unit 13 acquires the generated sentence based on the prompt generated by the generation unit 12 in step S12. A more specific description of the second acquisition unit 13 has been given above, and therefore will not be repeated here.

[0025] (Effect of information processing method S1) As described above, the information processing method S1 is configured to obtain an instruction content and reference information that is information to be referred to in the process indicated by the instruction content, generate a prompt by referring to the instruction content and the reference information, and obtain a generated sentence based on the prompt. Therefore, the information processing method S1 can provide a highly accurate business support technology.

[0026] Second Embodiment A second exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same functions as those described in the above exemplary embodiment will be given the same reference numerals, and their description will be omitted as appropriate. The scope of application of each technical means adopted in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technical means adopted in this exemplary embodiment can be adopted in other exemplary embodiments included in this disclosure, as long as no particular technical hindrance occurs. In addition, each technical means shown in each drawing referred to for explaining this exemplary embodiment can be adopted in other exemplary embodiments included in this disclosure, as long as no particular technical hindrance occurs.

[0027] (Configuration of information processing system 1A) The configuration of an information processing system 1A according to this exemplary embodiment will be described with reference to Fig. 3. Fig. 3 is a block diagram showing the configuration of the information processing system 1A. As shown in Fig. 3, the information processing system 1A includes an information processing device 100 and a generating device 50 connected to the information processing device 100 via a network N. Here, the specific configuration of the network N does not limit this exemplary embodiment, but as an example, a wireless LAN (Local Area Network), a wired LAN, a WAN (Wide Area Network), a public line network, a mobile data communication network, or a combination of these networks can be used.

[0028] (Generation device 50) As shown in FIG. 3, the generation device 50 includes a control unit 51, a storage unit 52, and a communication unit 53. The communication unit 53 communicates with a device external to the generation device 50. As an example, the communication unit 53 communicates with an information processing device 100 included in the information processing system 1A. The communication unit 53 transmits data supplied from the control unit 51 to the information processing device 100, and supplies data received from the information processing device 100 to the control unit 51. Note that the data received by the communication unit 53 from the information processing device 100 may include a prompt generated by the information processing device 100. In addition, the data provided by the communication unit 53 to the information processing device 100 may include a generated sentence generated by a language model described later based on the prompt.

[0029] A language model LM is stored in the storage unit 52. As an example, a plurality of parameters that define the language model LM are stored in the storage unit 52. As an example, these parameters are parameters that have been learned in advance by machine learning (parameters that have undergone an update process by machine learning), but this does not limit the present exemplary embodiment.

[0030] The control unit 51 executes the language model to obtain an output result from the language model LM. As an example, the control unit 51 inputs a prompt received from the information processing device 100 to the language model LM, and obtains a generated sentence generated by the language model LM. The control unit 51 also provides the generated sentence to the information processing device 100 via the communication unit 53.

[0031] In the present exemplary embodiment, the generation device 50 is illustrated as a device separate from the information processing device 100, but this does not limit the present exemplary embodiment. The control unit 51 included in the generation device 50 or the function of the language model execution unit in the control unit 51 may be configured to be included in the control unit of the information processing device 100. Similarly, the language model LM stored in the storage unit 52 included in the generation device 50 may be stored in the storage unit of the information processing device 100, and the information processing device 100 itself may be configured to be able to execute the language model LM.

[0032] (Configuration of information processing device 100) Next, the configuration of the information processing device 100 according to this exemplary embodiment will be described with reference to Fig. 3. As shown in Fig. 3, the information processing device 100 includes a control unit 10, a storage unit 20, a communication unit 30, and an input / output unit 40.

[0033] (Communication unit 30) The communication unit 30 communicates with devices external to the information processing device 100. As an example, the communication unit 30 communicates with the generation device 50. The communication unit 53 transmits data supplied from the control unit 10 to the generation device 50, and supplies data received from the generation device 50 to the control unit 10. Note that the data transmitted by the communication unit 30 to the generation device 50 may include a prompt generated by the control unit 10. Furthermore, the data received by the communication unit 30 from the generation device 50 may include a generated sentence generated by the generation device 50 based on the prompt.

[0034] (Input / output section 40) The input / output unit 40 is configured to include at least one of an input / output device such as a keyboard, a mouse, a display, a printer, a touch panel, etc. Alternatively, the input / output unit 40 may be configured to have an input / output device such as a keyboard, a mouse, a display, a printer, a touch panel, etc. connected to it. In this configuration, the input / output unit 40 accepts various types of information input to the information processing device 100 from the connected input device. Moreover, the input / output unit 40 outputs various types of information to the connected output device under the control of the control unit 10. An example of the input / output unit 40 is an interface such as a Universal Serial Bus (USB).

[0035] (Storage unit 20) The storage unit 20 stores various data referenced by the control unit 10 and various data generated by the control unit 10. As an example, the storage unit 20 stores: The user instruction IN acquired by the acquisition unit 110 via the input / output unit 40 or the communication unit 30 The input data IND acquired by the acquisition unit 110 via the input / output unit 40 or the communication unit 30 Instruction contents INC generated by the generation unit 12 with reference to at least one of the user instructions IN and the input data IND The reference information RI acquired by the acquisition unit 110 via the input / output unit 40 or the communication unit 30 The prompt PR generated by the control unit 10 with reference to the instruction content INC and the reference information RI A generated sentence GS generated by the generator 50 based on the prompt PR ·Multiple industry-specific models SM The various data stored in the storage unit 20 will be specifically described later.

[0036] (Control unit 10) 3, the control unit 10 includes a first acquisition unit 11, a second acquisition unit 13, and a generation unit 12, similar to the first exemplary embodiment. In this exemplary embodiment, the first acquisition unit 11 and the second acquisition unit 13 are collectively referred to as an acquisition unit 110.

[0037] (First Acquisition Unit 11) The first acquisition unit 11 acquires a user instruction IN via the input / output unit 40 or the communication unit 30, and stores the acquired user instruction IN in the storage unit 20. For example, the user instruction IN may include What kind of input data do you want to process? The user instruction acquired by the first acquisition unit 11 is referred to by the generation unit 12, which will be described later.

[0038] Moreover, the first acquisition unit 11 acquires input data IND accompanying the user instruction IN. The first acquisition unit 11 may acquire the input data IND together with the user instruction IN via the input / output unit 40 or the communication unit 30. Alternatively, the first acquisition unit 11 may acquire Refer to the acquired user instruction IN to identify input data to be used for executing a process based on the user instruction IN, The specified input data IND is obtained from the storage unit 20. The above configuration may also be used.

[0039] Further, the first acquisition unit 11 refers to at least one of the user instruction IN and the input data IND to identify the instruction content INC. As an example, the first acquisition unit 11 identifies content including both the user instruction IN and the input data IND as the instruction content INC.

[0040] Furthermore, the first acquisition unit 11 acquires reference information RI, which is information to be referenced in the process indicated by the instruction content INC. As described in the first exemplary embodiment, the reference information RI may be, for example, information selected so as to improve the accuracy of the sentence generated by the language model that references the prompt generated by the generation unit 12 described below. However, this does not limit this exemplary embodiment. Furthermore, the specific contents of the instruction content INC and the reference information RI do not limit this exemplary embodiment, but as an example, The instructions INC include an instruction to evaluate the evaluation target, The reference information RI includes past evaluation information, which is information indicating past evaluations of the evaluation target. Here, the evaluation target may or may not be configured to include at least a part of the above-mentioned input data IND.

[0041] (Generation unit 12) The generation unit 12 generates a prompt by referring to the instruction content INC and the reference information RI acquired (specified) by the first acquisition unit 11. Here, the prompt PR generated by the generation unit 12 is a prompt to be input to the language model LM executed by the generation device 50. Note that the specific configuration of the prompt generated by the generation unit 12 is not limited to this exemplary embodiment, but as an example, the generation unit 12 may be configured to generate a prompt PR including the instruction content INC and the reference information RI.

[0042] Also, as an example, the generation unit 12 may be configured to divide the above-mentioned evaluation target into a plurality of parts and include the evaluation target for each part in the prompt PR. Furthermore, the generation unit 12 may use a language model (called an industry-specific model) SM trained for each business category in the division process or the above-mentioned prompt generation process including the division process.

[0043] Here, the industry-specific model SM includes, as an example, A first industry-specific model SM1 trained by machine learning specialized for a first industry (first business category) A second industry-specific model SM2 trained by machine learning specialized for a second industry (second business category) A third industry-specific model SM3 trained by machine learning specialized for a third industry (third business category) More specifically, the industry-specific model SM includes: - Publishing industry email-specific model SM1 trained using training data containing multiple emails in the publishing industry - Financial industry-specific model SM2 trained using training data including multiple contracts or loan-related documents in the financial industry - General industry quotation specialized model SM3 trained using training data including multiple quotations in other general industries etc. As an example, the generation unit 12 can be configured to identify a target business category by referring to the user instruction IN acquired by the first acquisition unit 11, and execute the prompt generation process by using an industry-specific model that has been trained to be specialized for the identified business category. However, these examples do not limit the present exemplary embodiment.

[0044] The prompt PR generated by the generation unit 12 is provided to the generation device 50 via the communication unit 30, for example, and the generation device 50 uses the language model LM to generate a generated sentence GS based on the prompt PR.

[0045] (Second Acquisition Unit 13) The second acquiring unit 13 acquires the generated sentence GS generated by the language model LM based on the prompt PR, for example, via the communication unit 30. The acquired generated sentence GS is stored in the storage unit 20, for example.

[0046] (Processing example 1 by information processing device 100) Next, a first processing example by the information processing device 100 according to this exemplary embodiment will be described with reference to Fig. 4. Fig. 4 is a flow diagram showing a flow of the first processing example (also referred to as an information processing method S1A) by the information processing device 100.

[0047] (Step S101) In step S101, the first acquisition unit 11 acquires a user instruction IN and input data IND. -User instruction IN: "Please evaluate the business continuity of a certain company"; and As input data IND, at least a part of the business continuity plan of the company in question However, this example does not limit the present exemplary embodiment, and the first acquisition unit 11 may be configured to acquire any user instruction IN and any document corresponding to the user instruction IN in this step. The certain company or the business continuity plan can be expressed as a target for evaluation by the information processing device 100. The first acquisition unit 11 may also be configured to acquire templates or parts of templates of these target documents as input data IND.

[0048] In addition, in this step, the first acquisition unit 11 refers to at least one of the user instruction IN and the input data IND to identify (acquire) the instruction content INC. As an example, the first acquisition unit 11 identifies content including both the user instruction IN and the input data IND as the instruction content INC.

[0049] (Step S102) In step S102, the first acquisition unit 11 performs a process of dividing the input data IND acquired in step S101 into a plurality of parts (elements). As an example, the first acquisition unit 11 may divide a plurality of sentences or paragraphs included in the input data IND into a plurality of parts for each subject or viewpoint.

[0050] The first acquisition unit 11 may execute the division process using an industry-specific model SM that is trained to be specialized for a business category identified by referring to the user instruction IN. As described above, the industry-specific model is a model trained to be specialized for each business category, so that the input data IND can be divided with high accuracy by using the industry-specific model. The process of this step may be configured to be performed by the generation unit 12.

[0051] (Step S103) In step S103, the first acquisition unit 11 acquires reference information RI. Here, the reference information RI acquired by the first acquisition unit 11 in this step may be, for example, information selected so as to improve the accuracy of a generated sentence by a language model that references a prompt generated by the generation unit 12 described later. The reference information RI acquired by the first acquisition unit 11 in this step S103 may be, for example, - Past evaluations of the subject of evaluation (the "certain company" or "the business continuity plan of said certain company" mentioned above) It is.

[0052] The above-mentioned processes in steps S101 to S103 correspond to step S11 described in the first exemplary embodiment.

[0053] (Step S12) Next, in step S12, the generation unit 12 generates a prompt PR by referring to the instruction content INC acquired in step S101 and the reference information RI acquired in step S103. As an example, the generation unit 12 generates a prompt PR including the instruction content INC and the reference information RI.

[0054] 5 is a diagram showing a prompt PR1 which is an example of the prompt PR generated by the generation unit 12 in this step. As shown in FIG. 5, the prompt PR1 generated by the generation unit 12 is as follows: · Tags that indicate imperative statements (#imperative statements) · Command statement content (PR1_main) · Tags indicating constraints (#constraints) · Contents of constraints (INE11~INE13) · Tags showing past reviews (#past reviews) Past evaluations (RI11~RI12) · Tag indicating the input text (#input text) Input text content (EM11~EM12) · Tag indicating output statement (#output statement) Here, the command statement (PR1_main), constraint condition (INE11-INE13), and input statement (EM11-EM12) are, for example, the information contained in the instruction content INC described above expressed as each element of the prompt PR1.

[0055] More specifically, the command content (PR1_main) is the content generated by the generation unit 12 in response to the user instruction IN acquired by the first acquisition unit 11 in step S101. In the example shown in Fig. 5, the command content (PR1_main) includes the command "You are a professional loan officer. You are trying to evaluate and select a company that has excellent efforts in disaster prevention and business continuity. Please refer to the following "Past Evaluations" and output the evaluation results for the input sentence together with the reasons. The output should comply with the following constraints." corresponding to the user instruction IN "I would like you to evaluate the business continuity of a certain company" acquired in step S101.

[0056] On the other hand, the contents of the constraint conditions (INE11 to INE13) are, for example, contents generated by the generating unit 12 based on information included in the user instruction IN acquired by the first acquiring unit 11 in step S101. In the example shown in FIG. 5, the generating unit 12 generates each instruction element by referring to the user instruction IN. ·INE11: Please keep the length to less than 150 characters. ·INE12: Please clarify the evaluation perspective. ·INE13: Please indicate whether you like or dislike the rating. are used as the contents of the constraint conditions. However, this example does not limit the present exemplary embodiment, and predetermined contents may be used as the contents of the constraint conditions (INE11 to INE13).

[0057] The contents of the input sentence (EM11-EM12) indicate the input data IND acquired by the first acquisition unit 11 in step S101 and divided in step S102. More specifically, the contents of the input sentence (EM11-EM12) include elements EM11 and EM12 of the input sentence obtained by dividing the input data IND.

[0058] Here, as an example, the element EM11 is an element divided from the input data IND using the industry-specific model SM in step S102, and is an element having "compliance" as its theme or viewpoint. On the other hand, the element EM2 is an element divided from the input data IND using the industry-specific model SM in step S102, and is an element having "disaster prevention" as its theme or viewpoint.

[0059] The contents of the past evaluations (RI11 to RI12) are an example of the reference information RI acquired by the first acquisition unit 11 in step S103. In other words, the reference information RI includes past evaluation information that is information indicating past evaluations regarding the evaluation target. Here, in this example, the generation unit 12: Dividing the reference information RI so as to correspond to each element (or each subject or each viewpoint) of the input data IND divided in step S102; Each element of the reference information RI obtained by division is associated with each element of the input data IND and included in the prompt PR. In this way, the generation unit 12 may be configured to divide the evaluation target (input data IND) into a plurality of parts, and include the evaluation target for each part in the prompt PR. Here, the above-mentioned industry-specific model may be used for the division process of the reference information RI. Since the industry-specific model is a model that is trained specially for each business category, the use of the industry-specific model makes it possible to accurately divide the reference data RI.

[0060] More specifically, in the example shown in FIG. The reference information RI (past evaluation) is divided into elements RI11 and RI12 so as to correspond to each of the elements EM11 and EM12 of the input data IND, A prompt PR is generated that includes the divided elements RI11 and RI12.

[0061] In this step, the generation unit 12 adds the above-mentioned · Command statement content (PR1_main) · Contents of constraints (INE11~INE13) Past evaluations (RI11~RI12) Input text content (EM11~EM12) As an example, the generation unit 12 may generate the prompt PR by applying the following: · Tags that indicate imperative statements (#imperative statements) · Tags indicating constraints (#constraints) · Tags showing past reviews (#past reviews) · Tag indicating the input text (#input text) A prompt PR may be generated by applying the above-mentioned command statement content (PR1_main), constraint condition content (INE11 to INE13), past evaluations (RI11 to RI12), and input statement content (EM11 to EM12) to a prompt template including the above.

[0062] The prompt PR generated in this step is provided to the generation device 50, for example, via the communication unit 30. Then, in the generation device 50, the prompt is input to the language model LM, and the language model LM generates a generated sentence based on the prompt.

[0063] The control unit 10 may be configured to present the generated prompt PR to the user via the input / output unit 40, or may not be configured to present the prompt PR.

[0064] (Step S13) Next, in step S13, the second acquisition unit 13 acquires the generated sentence GS generated by the language model LM based on the prompt PR generated in step S12. Fig. 6 shows a generated sentence GS1, which is an example of the generated sentence GS acquired by the second acquisition unit 13 in this step.

[0065] In the example shown in Fig. 6, generated sentence GS1 includes elements GSE11 to GSE13. Here, in this example, element GSE11 indicates the evaluation result of the evaluation target ("Evaluation result: Highly rated" in the example of Fig. 6). Meanwhile, elements GSE12 to GSE13 indicate the grounds and reasons for generating the evaluation result, matters to be noted in the evaluation, etc., for each subject or viewpoint.

[0066] As an example, element GSE11 "Regarding compliance, ...(omitted)... is highly rated" indicates the evaluation reason from the viewpoint of "compliance," and element GSE12 "Regarding disaster prevention, ...(omitted)... is highly rated. However, ...(omitted)... efforts need to be strengthened" indicates the evaluation reason from the viewpoint of "disaster prevention" together with points to note. In this way, generated sentence GS1 contains the evaluation result for the evaluation target, which is the evaluation result from one or more viewpoints determined by referring to the past evaluation information.

[0067] (Step S104) Next, in step S104, the control unit 10 outputs the generated sentence GS acquired in step S13 as data via the input / output unit 40 or presents it to the user, for example.

[0068] (Step S105) Next, in step S105, The evaluation result indicated by the generated sentence GS obtained in step S13, and The evaluation result obtained by applying corrections by the user to the evaluation result indicated by the generated sentence GS obtained in step S13 At least one of the above is stored in the storage unit 20.

[0069] These evaluation results are stored in the storage unit 20 and are obtained in step S103 as reference information when evaluating new evaluation results.

[0070] The information processing device 100 according to this example has the same effects as the information processing device 1 according to the exemplary embodiment 1. In addition, in the information processing device 100 according to this example, The reference information RI includes past evaluation information (RI11-RI12), which is information showing past evaluations of the evaluation target (input data IND). The generated sentence GS includes evaluation results (GSE12 to GSE13) regarding the evaluation target, which are determined from one or more viewpoints by referring to the past evaluation information.

[0071] Therefore, according to the information processing device 100 of this example, it is possible to obtain an evaluation result with high accuracy.

[0072] (Additional notes regarding this processing example 1) In the present processing example 1, the processing may be performed as follows. The division process described in step S102 is performed in advance on multiple input data IND candidates, In step S102, an element corresponding to the input data IND obtained in step S101 is selected from the multiple elements previously divided by the division process. Such a configuration corresponds to executing the process in step S102 as pre-processing prior to step S101.

[0073] (Processing example 2 by information processing device 100) Next, a processing example 2 by the information processing device 100 according to this exemplary embodiment will be described with reference to Fig. 7. Fig. 7 is a flow diagram showing a flow of processing example 2 (also referred to as an information processing method S2A) by the information processing device 100.

[0074] (Step S201) In step S201, the first acquisition unit 11 acquires a user instruction IN and input data IND. -User instruction IN: "Please conduct a risk analysis of a certain company"; and As input data IND, at least a part of the business plan or securities report of the company in question However, this example does not limit the present exemplary embodiment, and the first acquisition unit 11 may be configured to acquire any user instruction IN and any document corresponding to the user instruction IN in this step. The certain company, the business plan, and the securities report can be expressed as being analysis targets by the information processing device 100. The first acquisition unit 11 may also be configured to acquire templates or parts of templates of these target documents as input data IND.

[0075] In addition, in this step, the first acquisition unit 11 refers to at least one of the user instruction IN and the input data IND to identify the instruction content INC. As an example, the first acquisition unit 11 identifies content including both the user instruction IN and the input data IND as the instruction content INC.

[0076] (Step S202) In step S202, the generation unit 12 determines an analysis viewpoint in the above analysis. As an example, the generation unit 12 may be configured to determine the analysis viewpoint by referring to the user instruction IN and input data IND acquired in step S201. The generation unit 12 may also be configured to use one or more predetermined analysis viewpoints. The generation unit 12 may also be configured to select one or more analysis viewpoints in the analysis from a plurality of predetermined analysis viewpoint candidates by referring to the user instruction IN and input data IND acquired in step S201.

[0077] In this way, the first acquisition unit 11 acquires information (also referred to as a second type of perspective information) indicating one or more analytical perspectives desired by the user in the analysis process, and the generation unit 12 may be configured to refer to the second type of perspective information to determine the perspective information (also referred to as a first type of perspective information) to be included in the prompt PR2 described later.

[0078] As an example, the generation unit 12 determines "credit risk" and "operational risk" as the analysis perspectives in the analysis. Note that in this specification, "operational risk" is also referred to as "op risk." The analysis perspectives determined in this step constitute an example of reference information RI in this example.

[0079] The above-mentioned processing in steps S201 and S202 corresponds to step S11 described in the first exemplary embodiment.

[0080] (Step S12) Next, in step S12, the generation unit 12 generates a prompt PR by referring to the instruction content INC acquired in step S201 and the reference information RI acquired in step S202. As an example, the generation unit 12 generates a prompt PR including the instruction content INC and the reference information RI.

[0081] 8 is a diagram showing a prompt PR2 which is an example of the prompt PR generated by the generation unit 12 in this step. As shown in FIG. 8, the prompt PR2 generated by the generation unit 12 is as follows: · Tags that indicate imperative statements (#imperative statements) - Command statement content (PR2_main) · Tags indicating constraints (#constraints) · Contents of constraints (INE21~INE23) · Tag indicating the output format (#output format) Output format contents (INE24) · Tags that indicate risk perspectives (#risk perspectives) Risk perspective content (RI21~RI22) · Tag indicating the input text (#input text) Input text content (EM21~EM23) · Tag indicating output statement (#output statement) Here, the command statement (PR2_main), constraint condition contents (INE21 to INE23), and input statement contents (EM21 to EM23) are, for example, each piece of information contained in the instruction content INC described above expressed as each element of the prompt PR2.

[0082] More specifically, the content of the command sentence (PR2_main) is the content generated by the generation unit 12 in response to the user instruction IN acquired by the first acquisition unit 11 in step S201. In the example shown in Fig. 8, the content of the command sentence (PR2_main) includes the command sentence "You are a professional loan officer. Please extract lines related to each risk viewpoint from the input sentence under the following constraint conditions" corresponding to the user instruction IN "I would like you to perform a risk analysis of a certain company" acquired in step S201.

[0083] On the other hand, the contents of the constraint conditions (INE21 to INE23) are, for example, contents generated by the generating unit 12 based on information included in the user instruction IN acquired by the first acquiring unit 11 in step S201. In the example shown in FIG. 8, the generating unit 12 generates each instruction element by referring to the user instruction IN. INE21: Please answer according to the output format below. INE12: If it spans multiple lines, write it in the format "start line~end line". INE13: Approximately 100 characters are used as the contents of the constraint conditions. However, this example does not limit the present exemplary embodiment, and predetermined contents may be used as the contents of the constraint conditions (INE21 to INE23).

[0084] The content of the output format (INE24) specifies the format of the generated sentence GS that is generated by referring to the prompt PR2. In this example, the content of the output format (INE24) is as follows: Please output in table format Along with a statement specifying a table format as the output format, ·Number of lines|Risk perspective|Rationale The text includes a statement specifying an item in a row or column of the table.

[0085] The contents of the risk perspectives (RI21 to RI22) are an example of the reference information RI (analysis perspectives in this example) acquired in step S202 by the first acquisition unit 11. In other words, the reference information RI includes information indicating one or more analysis perspectives in the analysis process (also referred to as first type of perspective information).

[0086] In the example shown in FIG. 8, the contents of the risk perspectives (RI21 to RI22) include, for example, Credit risk Operational Risk The document contains the following two analytical perspectives: Credit risk: The possibility that a security's... Operational risk: refers to the risk of administrative processing In this way, the reference information RI in this example includes, in addition to one or more analytical viewpoints, contents related to the explanation or definition of the viewpoints, so that a highly accurate generated sentence GS can be generated by the language model LM that refers to the reference information RI.

[0087] The contents of the input sentence (EM21 to EM23) indicate the input data IND acquired in step S101 by the first acquisition unit 11. As an example, the contents of the input sentence (EM21 to EM23) include element EM21 which is an element (sentence) on the first line included in the input data IND, element EM22 which is an element (sentence) on the second line, element EM23 which is an element (sentence) on the third line, ....

[0088] In this step, the generation unit 12 adds the above-mentioned - Command statement content (PR2_main) · Contents of constraints (INE21~INE23) Output format contents (INE24) Risk perspective content (RI21~RI22) Input text content (EM21~EM23) As an example, the generation unit 12 may generate the prompt PR by applying the following: · Tags that indicate imperative statements (#imperative statements) · Tags indicating constraints (#constraints) · Tag indicating the output format (#output format) · Tags that indicate risk perspectives (#risk perspectives) · Tag indicating the input text (#input text) A prompt PR may be generated by applying the above-mentioned command statement content (PR2_main), constraint condition content (INE21 to INE23), output format content (INE24), risk perspective content (RI21 to RI22), and input statement content (EM21 to EM23) to a prompt template including the above.

[0089] The prompt PR generated in this step is provided to the generation device 50, for example, via the communication unit 30. Then, in the generation device 50, the prompt is input to the language model LM, and the language model LM generates a generated sentence based on the prompt.

[0090] The control unit 10 may be configured to present the generated prompt PR to the user via the input / output unit 40, or may not be configured to present the prompt PR.

[0091] (Step S13) Next, in step S13, the second acquisition unit 13 acquires the generated sentence GS generated by the language model LM based on the prompt PR generated in step S12. Fig. 9 shows a generated sentence GS2, which is an example of the generated sentence GS acquired by the second acquisition unit 13 in this step.

[0092] In the example shown in FIG. 9, the generated sentence GS2 is configured in a table format in accordance with the contents of the output format included in the above-mentioned prompt PR2. Moreover, the generated sentence GS2 includes elements GSE21 to GSE22 as elements of each row. Moreover, each element includes a number indicating the "number of rows", a type of "risk perspective", and a "reason". Each of these items is in accordance with the contents of the output format included in the above-mentioned prompt PR2. Note that the above-mentioned "risk perspective" indicates a "risk perspective" regarding a risk determined to exist by the language model LM with reference to the above-mentioned prompt PR2.

[0093] More specifically, element GSE21 indicates in the third line of the input that a determination was made that credit risk exists. Element GSE21 also includes, as a basis or explanation for the determination, "The preparation and fairness of the financial statements can only provide a reasonable assurance as to their presentation, and may cease to function..."

[0094] On the other hand, element GSE22 indicates that it was determined that operational risk exists in lines 11 and 12 of the input text. Furthermore, element GSE22 includes, as the basis or explanation for the determination, "It is not possible to deal with changes in the business environment or atypical transactions that were not anticipated at the time the internal control system was established, and... (omitted)... the possibility cannot be denied."

[0095] In this way, the generated sentence GS2 includes the analysis results on the analysis target from each of one or more perspectives indicated by the first type of perspective information.

[0096] (Step S203) Subsequently, in step S203, the control unit 10 outputs the generated sentence GS acquired in step S13 as data via the input / output unit 40 or presents it to the user, for example.

[0097] The information processing device 100 according to this example has the same effects as the information processing device 1 according to the exemplary embodiment 1. In addition, in the information processing device 100 according to this example, The reference information RI includes first-type viewpoint information indicating one or more analysis viewpoints in the analysis process; The generated sentence GS2 includes the analysis results on the analysis target from each of one or more perspectives indicated by the first type of perspective information.

[0098] Therefore, according to the information processing device 100 of this example, it is possible to obtain highly accurate analysis results.

[0099] (Additional notes regarding this processing example 2) In addition, in the present processing example 2, the following processing may be further performed. In step S12 described above, the generation unit 12 includes one or more sentences that are less related to or unrelated to the content of the risk perspective in the content of the input sentence (EM21, . . . ), In the above-mentioned step S13, it is determined whether or not the content of the risk perspective in the acquired generated sentence GS2 includes content that is less relevant, If it is determined that the content of the risk perspective in the generated sentence GS2 is less relevant, the configuration content of the prompt PR2 is changed and a generated sentence GS2 based on the changed prompt PR2 is obtained.

[0100] With this configuration, it is possible to check whether the language model LM that references the prompt PR2 is performing highly accurate processing, and it is also possible to ensure that the information processing device 100 outputs a highly accurate generated sentence GS2.

[0101] In addition, the determination process in step S13 may be configured to use the above-mentioned industry-specific model, as an example. The industry-specific model is a model that is trained to be specialized for the business category, so by using the industry-specific model, it is possible to accurately eliminate low-accuracy output results by the language model LM.

[0102] (Processing example 3 by information processing device 100) Next, a processing example 3 by the information processing device 100 according to this exemplary embodiment will be described with reference to Fig. 10. Fig. 7 is a flow diagram showing a flow of processing example 3 (also referred to as an information processing method S3A) by the information processing device 100.

[0103] (Step S301) In step S301, the first acquisition unit 11 acquires a user instruction IN and input data IND. - User instruction IN: "Please check a certain document" As input data IND, at least a part of the document Here, the specific example of the document does not limit the present exemplary embodiment, but may be, for example, an application document prepared by an applicant, a request for approval, etc. The document can be expressed as a confirmation target (check target) by the information processing device 100. The first acquisition unit 11 may also be configured to acquire, as the input data IND, a template of the target document or a part of the template.

[0104] In addition, in this step, the first acquisition unit 11 refers to at least one of the user instruction IN and the input data IND to identify the instruction content INC. As an example, the first acquisition unit 11 identifies content including both the user instruction IN and the input data IND as the instruction content INC.

[0105] (Step S302) In step S302, the generation unit 12 acquires comparison information to be compared with the document in the above confirmation (check). Here, the generation unit 12 may be configured to use one or more pieces of comparison information that are predetermined. The generation unit 12 may also be configured to select one or more pieces of comparison information for the confirmation (check) from a plurality of predetermined candidates of comparison information by referring to the user instruction IN and input data IND acquired in step S301. The comparison information acquired in this step constitutes an example of reference information RI in this example.

[0106] The above-mentioned processing in steps S301 and S302 corresponds to step S11 described in the first exemplary embodiment.

[0107] (Step S12) Next, in step S12, the generation unit 12 generates a prompt PR by referring to the instruction content INC acquired in step S301 and the reference information RI acquired in step S302. As an example, the generation unit 12 generates a prompt PR including the instruction content INC and the reference information RI.

[0108] 11 is a diagram showing a prompt PR3 which is an example of the prompt PR generated by the generation unit 12 in this step. As shown in FIG. 11, the prompt PR3 generated by the generation unit 12 is as follows: · Tags that indicate imperative statements (#imperative statements) - Command statement content (PR3_main) · Tags that indicate master data (#masterdata) · Master data contents (EM31~EM33) Tag indicating control data (#control data) Contents of control data (RI31~RI32) · Tag indicating output statement (#output statement) Here, the command statement content (PR3_main) and the master data content (EM31 to EM33) are, for example, each piece of information contained in the instruction content INC described above expressed as each element of the prompt PR3.

[0109] More specifically, the content of the command sentence (PR3_main) is the content generated by the generating unit 12 in response to the user instruction IN acquired by the first acquiring unit 11 in step S301. In the example shown in Fig. 11, the content of the command sentence (PR3_main) contains the command sentence "You are a professional loan officer. Please compare the master data below with the control data and list the inconsistencies" corresponding to the user instruction IN "I would like you to confirm (check) a certain document" acquired in step S301.

[0110] In the above example, the content of the command statement (PR3_main) can also be expressed as including an instruction to include the differences in the generated statement GS3 (described later) if there is a difference between the target to be checked (the document mentioned above) and the comparison information (the comparison data mentioned above).

[0111] The contents of the master data (EM31 to EM33) indicate the input data IND acquired by the first acquisition unit 11 in step S101. As an example, the contents of the master data (EM31 to EM33) include, as definition statements of each component included in the master data (input data IND), ·EM31: The following data represents the element ID, context ID,... (omitted)... value. The specific numerical data for each component is as follows: ·EM32:Net income per share,···(omitted)···,301.71 ·EM33: Number of employees,···(omitted)···,117000 In the above example, the master data is expressed in a data format equivalent to the CSV (Comma Separated Values) format, but this is not a limitation of the present example. As described above, by including definition statements of each component included in the master data in the contents of the master data, any data format can be suitably used as the master data.

[0112] The contents of the comparison data (RI31 to RI32) are an example of the reference information RI (comparison information in this example) acquired in step S302 by the first acquisition unit 11. In other words, the reference information RI includes comparison information that is information to be compared with the check target in the confirmation (check) process.

[0113] In the example shown in FIG. 11, the contents of the comparison data (RI31 to RI32) are the following definition statements for each component included in the comparison data: ·RI31: The following data shows the segment name and number of employees. The specific numerical data for each component is as follows: Segment A: 5100 Segment B: 16200 ...(snip)... Total: 117200 In the above example, the control data is also expressed in a data format corresponding to the CSV format, but this is not a limitation of this example. As described above, by including definitions of each component included in the control data in the contents of the control data, any data format can be suitably used as the control data.

[0114] In this step, the generation unit 12 adds the above-mentioned - Command statement content (PR3_main) Master data contents (INE31~INE33) Contents of control data (RI31~RI32) As an example, the generation unit 12 may generate the prompt PR by applying the following: · Tags that indicate imperative statements (#imperative statements) · Tags that indicate master data (#masterdata) Tag indicating control data (#control data) The prompt PR may be generated by applying the above-mentioned command statement content (PR3_main), master data content (INE31 to INE33), and comparison data content (RI31 to RI32) to a prompt template including the above.

[0115] The prompt PR generated in this step is provided to the generation device 50, for example, via the communication unit 30. Then, in the generation device 50, the prompt is input to the language model LM, and the language model LM generates a generated sentence based on the prompt.

[0116] The control unit 10 may be configured to present the generated prompt PR to the user via the input / output unit 40, or may not be configured to present the prompt PR.

[0117] (Step S13) Next, in step S13, the second acquisition unit 13 acquires the generated sentence GS generated by the language model LM based on the prompt PR generated in step S12. Fig. 12 shows a generated sentence GS3, which is an example of the generated sentence GS acquired by the second acquisition unit 13 in this step.

[0118] In the example shown in FIG. 12, the generated sentence GS3 includes elements GSE31 to GSE32. The element GSE31 is · There is a difference between the master data and the control data · Items that differ between master data and control data Specifically, the element GSE31 contains the content "The total number of employees in the control data is different from the number of employees in the master data." In this way, the generated sentence GS3 contains information that indicates the differences between the master data and the control data.

[0119] On the other hand, element GSE32 includes an explanation or justification for the difference between the master data and the control data. For example, in the example shown in Fig. 12, element GSE32 includes the content "The number of employees in the master data is... (omitted)..., and the number of employees in the control data is... (omitted)..., resulting in a contradiction."

[0120] (Step S303) Next, in step S303, the control unit 10 outputs the generated sentence GS acquired in step S13 as data via the input / output unit 40 or presents it to the user, for example.

[0121] The information processing device 100 according to this example has the same effects as the information processing device 1 according to the exemplary embodiment 1. In addition, in the information processing device 100 according to this example, The reference information RI includes comparison information, which is information to be compared with the check target in the check process; The instruction content INC in the prompt PR3 (for example, the command PR3_main) includes an instruction to include the difference in the generated sentence if the check target and the comparison information are different. The generated sentence GS3 includes information indicating the above-mentioned differences.

[0122] Therefore, according to the information processing device 100 of this example, it is possible to obtain a highly accurate confirmation (check) result.

[0123] Third embodiment A third exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same functions as those described in the above exemplary embodiment will be given the same reference numerals, and their description will be omitted as appropriate. The scope of application of each technical means employed in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technical means employed in this exemplary embodiment can be employed in other exemplary embodiments included in this disclosure, as long as no particular technical hindrance occurs. In addition, each technical means shown in each drawing referred to for explaining this exemplary embodiment can be employed in other exemplary embodiments included in this disclosure, as long as no particular technical hindrance occurs.

[0124] (Configuration of information processing system 1B) The configuration of an information processing system 1B according to this exemplary embodiment will be described with reference to Fig. 13. Fig. 13 is a block diagram showing the configuration of the information processing system 1B. As shown in Fig. 13, the information processing system 1B includes an information processing device 200 and a generating device 50 connected to the information processing device 200 via a network N. Here, the specific configuration of the network N does not limit this exemplary embodiment, but as an example, a wireless LAN (Local Area Network), a wired LAN, a WAN (Wide Area Network), a public line network, a mobile data communication network, or a combination of these networks can be used.

[0125] (Generation device 50) The generating device 50 is similar to that described in the second exemplary embodiment, and therefore a description thereof will be omitted here.

[0126] (Information processing device 200) The information processing device 200 according to this exemplary embodiment includes a specification unit 14 and a correction unit 15 in addition to the components included in the information processing device 100 according to exemplary embodiment 2. The following description will focus on the differences between the information processing device 200 and the information processing device 100.

[0127] (Specific part 14) The identification unit 14 identifies a target business category and provides category information indicating the identified business category to the generation unit 12. As an example, the identification unit 14 identifies the target business category by referring to a wording included in the user instruction IN acquired by the first acquisition unit 11. Alternatively, the identification unit 14 may be configured to acquire information indicating a user's selection regarding a business category and identify the target business category by referring to the information.

[0128] Here, the "business category" in this exemplary embodiment refers to a category related to at least one of an industry and a job type (job), for example. For example, the business category may include: ·Publishing industry ·Financial industry Other general industries In addition, each category may include, for example, Email Creation Title suggestions Needs survey · Estimate preparation · Contract drafting · Legal check of contracts Here, a certain subcategory may exist across multiple categories. For example, a subcategory called "email creation" may exist in any of the categories of the publishing industry, the finance industry, and other general industries. In this exemplary embodiment, the term "business category" may include the above-mentioned "industry category" and "job subcategory," but this is not intended to limit this exemplary embodiment.

[0129] In addition, the specification unit 14 Generate a prompt for identifying an industry, the prompt including a user instruction IN and an instruction sentence for identifying a business category of the user instruction IN; The generated industry-specific prompt is input to a language model LM included in the generation device 50; - Identify the target business category by referring to the output of the language model LM The following process may be performed.

[0130] As another example, the determination unit 14 · Input user-specified IN to multiple industry-specific models, and identify the target business category by referring to the output from each industry-specific model. As described above, the industry-specific model is a model that is trained to be specialized for each business category, and therefore, by using the industry-specific model, it is possible to accurately identify the target business category.

[0131] Alternatively, the specification unit 14 may be configured to refer to a dictionary (keyword table) for each business category and specify the target business category depending on whether or not the user instruction IN contains a word in the dictionary.

[0132] (Generation unit 12) The generating unit 12 according to this exemplary embodiment refers to category information indicating the target task category in the process performed by the generating unit 12 according to the exemplary embodiment 2. As an example, in the process of generating the prompt PR in step S12 described above, the generating unit 12 refers to category information indicating the target task category and generates a prompt PR according to the target task category. As an example, - Generate the contents of the command sentences included in the prompt PR so that they correspond to the business category of the target. From a plurality of elements (words, sentences, paragraphs, etc.) included in the input data IND, elements corresponding to the target business category are extracted, and the extracted elements are used to generate a prompt PR (for this process, an industry-specific model specialized for the target business category may be used). From the multiple elements (words, sentences, paragraphs, etc.) included in the reference information RI, elements corresponding to the target business category are extracted, and the extracted elements are used to generate the prompt PR (for this process, an industry-specific model specialized for the target business category may be used). In addition, the generation unit 12 may perform other processes according to the target business category in the above-mentioned other processes.

[0133] (revised part 15) The modification unit 15 modifies the generated sentences GS generated by the generation device 50. More specifically, the modification unit 15 modifies the generated sentences GS in accordance with the business category identified by the identification unit 14. As an example, the modification unit 15 may be configured to modify the generated sentences GS by using the above-mentioned business type specialization model SM.

[0134] For example, the correction unit 15 may correct an expression that is included in the generated sentence GS and is inappropriate for the business category identified by the identification unit 14, to an expression that is appropriate for the business category, using an industry-specific model that is learned specifically for the business category identified by the identification unit 14. The corrected generated sentence RGS may be configured to be provided to the user via the input / output unit 40 or the communication unit 30, for example.

[0135] (Flow of information processing method S1B) Next, the flow of the information processing method S1B according to this exemplary embodiment will be described with reference to Fig. 14. Fig. 14 is a flow diagram showing the flow of the information processing method S1B executed by the information processing device 200 according to this exemplary embodiment.

[0136] (Step S401) In step S401, the first acquisition unit 11 acquires a user instruction IN and input data IND. Then, the first acquisition unit 11 identifies an instruction content INC from the user instruction IN and the input data IND. As an example, in this step, the first acquisition unit 11 performs at least one of the processes of steps S11, S101, S201, and S301 described above.

[0137] (Step S402) In step S402, the identification unit 14 identifies the target business category. As an example, the identification unit 14 identifies the target business category by referring to at least one of the user instruction IN and the input data IND acquired in step S401. The specific process by the identification unit 14 has been described above, and therefore will not be described here.

[0138] (Step S403) In step S403, the first acquisition unit 11 acquires reference information RI. As an example, in this step, the first acquisition unit 11 performs at least one of the processes of steps S11, S103, S202, and S302 described above. In addition, in this step, the first acquisition unit 11 may be configured to select and acquire one or more reference information RI from a plurality of reference information RI candidates according to the target business category.

[0139] The above steps S401 to S403 correspond to step S11 described in the first exemplary embodiment.

[0140] (Step S12) Next, in step S12, the generation unit 12 generates a prompt PR by referring to the instruction content INC acquired in step S401 and the reference information RI acquired in step S403. The process of the generation unit 12 in this step includes the same process as in step S12 described above. Also, in this step, the generation unit 12 refers to the task category identified in step S402 and generates a prompt PR according to the task category. As an example, the generation unit 12 - Generate the contents of the command sentences included in the prompt PR so that they correspond to the business category of the target. From a plurality of elements (words, sentences, paragraphs, etc.) included in the input data IND, elements corresponding to the target business category are extracted, and the extracted elements are used to generate a prompt PR (for this process, an industry-specific model specialized for the target business category may be used). From the multiple elements (words, sentences, paragraphs, etc.) included in the reference information RI, elements corresponding to the target business category are extracted, and the extracted elements are used to generate the prompt PR (for this process, an industry-specific model specialized for the target business category may be used). The above processing may be performed.

[0141] (Step S13) Next, in step S13, the second acquisition unit 13 acquires the generated sentence GS generated by the language model LM based on the prompt PR generated in step S12. Examples of the generated sentence acquired in this step include the examples described in the above-described exemplary embodiment.

[0142] (Step S404) Next, in step S404, the correction unit 15 corrects the generated sentence GS acquired in step S13. As an example, the correction unit 15 corrects the generated sentence GS according to the business category characterized in step S402. For example, the correction unit 15 may be configured to correct the generated sentence GS using an industry-specific model SM that has been trained to specialize in the target business category identified in step S402. The specific processing by the correction unit 15 has been described above, and therefore will not be described here.

[0143] The information processing device 200 according to this example has the same effects as the information processing devices 1 and 100 according to the above-described exemplary embodiments. Identify the target business category; Generate prompt PRs according to the identified business categories, - Modify the generated sentence GS based on the prompt according to the identified task category. According to this configuration, it is possible to provide a business support technology with higher accuracy. [Software implementation example] A part or all of the functions of the information processing devices 1, 100, and 200 may be realized by hardware such as an integrated circuit (IC chip), or may be realized by software.

[0144] In the latter case, the information processing devices 1, 100, 200 are realized, for example, by a computer that executes instructions of a program, which is software that realizes each function. An example of such a computer (hereinafter, referred to as computer C) is shown in FIG. 15. The computer C includes at least one processor C1 and at least one memory C2. The memory C2 stores a program P for operating the computer C as the information processing device 1, 100, 200. In the computer C, the processor C1 reads and executes the program P from the memory C2, thereby realizing each function of the information processing device 1, 100, 200.

[0145] The processor C1 may be, for example, a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a tensor processing unit (TPU), a quantum processor, a microcontroller, or a combination thereof. The memory C2 may be, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination thereof.

[0146] The computer C may further include a RAM (Random Access Memory) for expanding the program P during execution and for temporarily storing various data. The computer C may further include a communication interface for transmitting and receiving data to and from other devices. The computer C may further include an input / output interface for connecting input / output devices such as a keyboard, a mouse, a display, and a printer.

[0147] Furthermore, the program P can be recorded on a non-transitory tangible recording medium M that can be read by the computer C. Such a recording medium M can be, for example, a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit. The computer C can obtain the program P via such a recording medium M. Furthermore, the program P can be transmitted via a transmission medium. Such a transmission medium can be, for example, a communication network or broadcast waves. The computer C can also obtain the program P via such a transmission medium.

[0148] [Additional Note 1] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0149] (Appendix A1) a first acquisition means for acquiring an instruction content and reference information which is information to be referenced in a process indicated by the instruction content; a generating means for generating a prompt by referring to the instruction content and the reference information; A second acquisition means for acquiring a generated sentence based on the prompt; An information processing device comprising:

[0150] (Appendix A2) The generating means generates a prompt including the instruction content and the reference information. 2. The information processing device according to claim 1 .

[0151] (Appendix A3) The instruction content includes an instruction to evaluate the evaluation target, The reference information includes past evaluation information that is information indicating past evaluations of the evaluation target. 13. The information processing device according to claim 12,

[0152] (Appendix A4) The generated sentence includes an evaluation result regarding the evaluation target, the evaluation result being determined from one or more viewpoints with reference to the past evaluation information. 2. The information processing device according to claim 1 ,

[0153] (Appendix A5) The generating means includes: Dividing the evaluation object into a plurality of parts; Include in the prompt what each part is evaluated for 13. The information processing device according to claim 12, wherein the information processing device is a

[0154] (Appendix A6) The instruction content includes an instruction to analyze the analysis target, The reference information includes first viewpoint information indicating one or more analytical viewpoints in the analysis process. 13. The information processing device according to claim 12,

[0155] (Appendix A7) The generated sentence includes an analysis result regarding the analysis target, the analysis result being from one or more perspectives indicated by the first type of perspective information. 2. The information processing device according to claim 1 ,

[0156] (Appendix A8) the first acquisition means acquires second type of viewpoint information indicating one or more analytical viewpoints desired by a user in the analysis process; The generating means determines the first type of viewpoint information by referring to the second type of viewpoint information. 13. The information processing device according to claim 12, wherein the information processing device is a

[0157] (Appendix A9) The instruction content includes an instruction to check a check target, The reference information includes comparison information that is information to be compared with the check target in the check process. 13. The information processing device according to claim 12,

[0158] (Appendix A10) The instructions include an instruction to include the difference in the generated sentence when the check target and the comparison information are different. 2. The information processing device according to claim 1 ,

[0159] (Appendix A11) The generated sentence includes information indicating the difference. 2. The information processing device according to claim 10 .

[0160] (Appendix A12) A means for identifying a target business category is further provided, The generating means generates a prompt according to the task category identified by the identifying means. 2. An information processing device according to claim 1,

[0161] (Appendix A13) The system further includes a correction means for correcting the generated sentence in accordance with the task category identified by the identification means. 2. The information processing device according to claim 12.

[0162] (Appendix B1) At least one processor Acquiring instruction content and reference information that is information to be referred to in a process indicated by the instruction content; generating a prompt by referencing the instruction content and the reference information; obtaining a generated sentence based on the prompt; An information processing method comprising the steps of:

[0163] (Appendix B2) The generating step generates a prompt including the instruction content and the reference information. 2. An information processing method as described in Appendix B1.

[0164] (Appendix B3) The instruction content includes an instruction to evaluate the evaluation target, The reference information includes past evaluation information that is information indicating past evaluations of the evaluation target. 13. An information processing method according to claim 12, wherein the information processing method is a method for processing information according to claim 13.

[0165] (Appendix B4) The generated sentence includes an evaluation result regarding the evaluation target, the evaluation result being determined from one or more viewpoints with reference to the past evaluation information. 2. An information processing method as described in Appendix B3.

[0166] (Appendix B5) The generating step includes: Dividing the evaluation object into a plurality of parts; Include in the prompt what each part is evaluated for 13. An information processing method according to claim 12, wherein said information processing method is

[0167] (Appendix B6) The instruction content includes an instruction to analyze the analysis target, The reference information includes first viewpoint information indicating one or more analytical viewpoints in the analysis process. 13. An information processing method according to claim 12, wherein the information processing method is a method for processing information according to claim 13.

[0168] (Appendix B7) The generated sentence includes an analysis result regarding the analysis target, the analysis result being from one or more perspectives indicated by the first type of perspective information. 2. An information processing method as described in Appendix B6.

[0169] (Appendix B8) the acquiring step acquires second type viewpoint information indicating one or more analysis viewpoints desired by a user in the analysis processing; The generating step determines the first type of viewpoint information by referring to the second type of viewpoint information. 2. An information processing method according to claim 1, wherein said information processing method is

[0170] (Appendix B9) The instruction content includes an instruction to check a check target, The reference information includes comparison information that is information to be compared with the check target in the check process. 13. An information processing method according to claim 12, wherein the information processing method is a method for processing information according to claim 13.

[0171] (Appendix B10) The instructions include an instruction to include the difference in the generated sentence when the check target and the comparison information are different. 2. An information processing method as described in Appendix B9.

[0172] (Appendix B11) The generated sentence includes information indicating the difference. 2. An information processing method as described in Appendix B10.

[0173] (Appendix B12) further including identifying a target business category; The generating step generates a prompt according to the task category identified in the identifying step. 13. An information processing method according to any one of appendices B1 to B11.

[0174] (Appendix B13) The method further includes modifying the generated sentence according to the task category identified in the identifying step. 2. An information processing method as described in Appendix B12.

[0175] (Appendix C1) On the computer, a first acquisition process for acquiring an instruction content and reference information that is information to be referenced in a process indicated by the instruction content; a generation process for generating a prompt by referring to the instruction content and the reference information; a second acquisition process for acquiring a generated sentence based on the prompt; A program that executes the following.

[0176] (Appendix C2) The generating process generates a prompt including the instruction content and the reference information. 2. The program described in Appendix C1.

[0177] (Appendix C3) The instruction content includes an instruction to evaluate the evaluation target, The reference information includes past evaluation information that is information indicating past evaluations of the evaluation target. 1. A program according to appendix C1 or C2.

[0178] (Appendix C4) The generated sentence includes an evaluation result regarding the evaluation target, the evaluation result being determined from one or more viewpoints with reference to the past evaluation information. 2. The program described in Appendix C3.

[0179] (Appendix C5) The generation process includes: Dividing the evaluation object into a plurality of parts; Include in the prompt what each part is evaluated for 2. A program as described in Appendix C3 or C4.

[0180] (Appendix C6) The instruction content includes an instruction to analyze the analysis target, The reference information includes first viewpoint information indicating one or more analytical viewpoints in the analysis process. 1. A program according to appendix C1 or C2.

[0181] (Appendix C7) The generated sentence includes an analysis result regarding the analysis target, the analysis result being from one or more perspectives indicated by the first type of perspective information. 2. The program described in Appendix C6.

[0182] (Appendix C8) the first acquisition process acquires second type of viewpoint information indicating one or more analysis viewpoints desired by a user in the analysis process; The generation process determines the first type of viewpoint information by referring to the second type of viewpoint information. 2. A program as described in Appendix C6 or C7.

[0183] (Appendix C9) The instruction content includes an instruction to check a check target, The reference information includes comparison information that is information to be compared with the check target in the check process. 1. A program according to appendix C1 or C2.

[0184] (Appendix C10) The instructions include an instruction to include the difference in the generated sentence when the check target and the comparison information are different. 2. The program described in Appendix C9.

[0185] (Appendix C11) The generated sentence includes information indicating the difference. 2. The program described in Appendix C10.

[0186] (Appendix C12) Further executing a specification process for specifying a target business category; The generating process generates a prompt according to the task category identified by the identifying process. 13. A program according to any one of appendices C1 to C11.

[0187] (Appendix C13) A correction process is further executed to correct the generated sentence according to the task category identified by the identification process. 2. The program described in Appendix C12.

[0188] [Additional Note 2] This disclosure includes the techniques described in the following appendices. However, the present invention is not limited to the techniques described in the following appendices, and various modifications are possible within the scope of the claims.

[0189] (Appendix D1) at least one processor, the at least one processor comprising: a first acquisition process for acquiring an instruction content and reference information that is information to be referenced in a process indicated by the instruction content; a generation process for generating a prompt by referring to the instruction content and the reference information; a second acquisition process for acquiring a generated sentence based on the prompt; An information processing device that executes the above.

[0190] (Appendix D2) The generating process generates a prompt including the instruction content and the reference information. 2. The information processing device according to claim 1 .

[0191] (Appendix D3) The instruction content includes an instruction to evaluate the evaluation target, The reference information includes past evaluation information that is information indicating past evaluations of the evaluation target. The information processing device according to claim D1 or D2.

[0192] (Appendix D4) The generated sentence includes an evaluation result regarding the evaluation target, the evaluation result being determined from one or more viewpoints with reference to the past evaluation information. 5. The information processing device according to claim 3,

[0193] (Appendix D5) The generation process includes: Dividing the evaluation object into a plurality of parts; Include in the prompt what each part is evaluated for 5. The information processing device according to claim 3 or 4.

[0194] (Appendix D6) The instruction content includes an instruction to analyze the analysis target, The reference information includes first viewpoint information indicating one or more analytical viewpoints in the analysis process. The information processing device according to claim D1 or D2.

[0195] (Appendix D7) The generated sentence includes an analysis result regarding the analysis target, the analysis result being from one or more perspectives indicated by the first type of perspective information. 13. The information processing device according to claim 12,

[0196] (Appendix D8) the first acquisition process acquires second type of viewpoint information indicating one or more analysis viewpoints desired by a user in the analysis process; The generation process determines the first type of viewpoint information by referring to the second type of viewpoint information. An information processing device according to appended claim D6 or D7.

[0197] (Appendix D9) The instruction content includes an instruction to check a check target, The reference information includes comparison information that is information to be compared with the check target in the check process. The information processing device according to claim D1 or D2.

[0198] (Appendix D10) The instructions include an instruction to include the difference in the generated sentence when the check target and the comparison information are different. 11. The information processing device according to claim 9,

[0199] (Appendix D11) The generated sentence includes information indicating the difference. 2. The information processing device according to claim 10 .

[0200] (Appendix D12) Further performing a process of identifying a target business category; The generating process generates a prompt according to the task category identified by the identifying process. 2. An information processing device according to claim 1, wherein the information processing device is a

[0201] (Appendix D13) A correction process is further executed to correct the generated sentence according to the task category identified by the identification process. 2. The information processing device according to claim 12. [Explanation of symbols]

[0202] 1,100,200 ···Information processing equipment 11 First acquisition unit (first acquisition means) 12...Generation unit (generation means) 13 Second acquisition unit (second acquisition means) 14 ... Specification part (specification means) 15... Correction section (correction means)

Claims

1. a first acquiring means for acquiring an instruction content and reference information which is information to be referred to in a process indicated by the instruction content; a generating means for generating a prompt by referring to the instruction content and the reference information; A second acquisition means for acquiring a generated sentence based on the prompt; An information processing device comprising:

2. The generating means generates a prompt including the instruction content and the reference information. The information processing device according to claim 1 .

3. The instruction content includes an instruction to evaluate the evaluation target, The reference information includes past evaluation information that is information indicating past evaluations of the evaluation target. The information processing device according to claim 2 .

4. The generated sentence includes an evaluation result regarding the evaluation target, the evaluation result being determined from one or more perspectives with reference to the past evaluation information. The information processing device according to claim 3 .

5. The generating means includes: Dividing the evaluation object into a plurality of parts; Include in the prompt what each part is evaluated for The information processing device according to claim 4.

6. The instruction content includes an instruction to analyze the analysis target, The reference information includes first type viewpoint information indicating one or more analysis viewpoints in the analysis process. The information processing device according to claim 2 .

7. The generated sentence includes an analysis result regarding the analysis target, the analysis result being from one or more perspectives indicated by the first type of perspective information. The information processing device according to claim 6.

8. the first acquisition means acquires second type of viewpoint information indicating one or more analysis viewpoints desired by a user in the analysis process; The generating means determines the first type of viewpoint information by referring to the second type of viewpoint information. The information processing device according to claim 7.

9. The instruction content includes an instruction to check a check target, The reference information includes comparison information that is information to be compared with the check target in the check process. The information processing device according to claim 2 .

10. The instructions include an instruction to include the difference in the generated sentence when the check target and the comparison information are different. The information processing device according to claim 9.

11. The generated sentence includes information indicating the difference. The information processing device according to claim 10.

12. A means for identifying a target business category is further provided, The generating means generates a prompt according to the task category identified by the identifying means. The information processing device according to claim 1 .

13. The system further includes a correction means for correcting the generated sentence in accordance with the task category identified by the identification means. The information processing device according to claim 12.

14. At least one processor Acquiring instruction content and reference information that is information to be referred to in a process indicated by the instruction content; generating a prompt by referencing the instruction content and the reference information; obtaining a generated sentence based on the prompt; An information processing method comprising the steps of:

15. On the computer, a first acquisition process for acquiring an instruction content and reference information that is information to be referenced in a process indicated by the instruction content; a generation process for generating a prompt by referring to the instruction content and the reference information; a second acquisition process for acquiring a generated sentence based on the prompt; A program that executes the following.

Images